Development of a prediction model to identify undiagnosed chronic obstructive pulmonary disease patients in primary care settings in China

Abstract

Abstract. Background:. At present, a large number of chronic obstructive pulmonary disease (COPD) patients are undiagnosed in China. Thus, this study aimed to develop a simple prediction model as a screening tool to identify patients at risk for COPD. Methods:. The study was based on the data of 22,943 subjects aged 30 to 79 years and enrolled in the second resurvey of China Kadoorie Biobank during 2012 and 2013 in China. We stepwisely selected the predictors using logistic regression model. Then we tested the model validity through P–P graph, area under the receiver operating characteristic curve (AUROC), ten-fold cross validation and an external validation in a sample of 3492 individuals from the Enjoying Breathing Program in China. Results:. The final prediction model involved 14 independent variables, including age, sex, location (urban/rural), region, educational background, smoking status, smoking amount (pack-years), years of exposure to air pollution by cooking fuel, family history of COPD, history of tuberculosis, body mass index, shortness of breath, sputum and wheeze. The model showed an area under curve (AUC) of 0.72 (95% confidence interval [CI]: 0.72–0.73) for detecting undiagnosed COPD patients, with the cutoff of predicted probability of COPD=0.22, presenting a sensitivity of 70.13% and a specificity of 62.25%. The AUROC value for screening undiagnosed patients with clinically significant COPD was 0.68 (95% CI: 0.66–0.69). Moreover, the ten-fold cross validation reported an AUC of 0.72 (95% CI: 0.71–0.73), and the external validation presented an AUC of 0.69 (95% CI: 0.68–0.71). Conclusion:. This prediction model can serve as a first-stage screening tool for undiagnosed COPD patients in primary care settings.